Skip to content

P
Paddle-Lite

PaddlePaddle / Paddle-Lite

通知 337
Star 4
Fork 1
P
Paddle-Lite
提交 64aa8f05 编写于 作者: Y Yanzhan Yang 提交者： GitHub
浏览文件
操作

shift filter conversion process to loading phase (#1606) 

* shift filter conversion process to loading phase

* fix winograd edge case
上级 b9ce0794
隐藏空白更改
内联 并排
Showing with 184 addition and 189 deletion
metal/paddle-mobile-metallib/paddle-mobile-metallib/ConvAddReluMetal.metal
浏览文件 @ 64aa8f05
......@@ -427,241 +427,115 @@ kernel void depthwise_conv_add_relu_3x3_half(texture2d_array<half, access::sampl
kernel void depthwise_conv_add_relu_3x3_half_winograd(texture2d_array<half, access::sample> inTexture [[texture(0)]],
texture2d_array<half, access::write> outTexture [[texture(1)]],
constant MetalConvParam &param [[buffer(0)]],
const device half *weights [[buffer(1)]],
const device half4x4 *weights [[buffer(1)]],
const device half4 *biase [[buffer(2)]],
uint3 gid [[thread_position_in_grid]]) {
uint x = gid.x, y = gid.y;
uint ow = outTexture.get_width();
if (ow % 2 != 0) {
ow++;
}
uint oh = outTexture.get_height();
if (gid.x >= ow || gid.y >= oh) {
if (oh % 2 != 0) {
oh++;
}
if (x >= ow || y >= oh) {
return;
}
uint tx = (gid.x / 2) * 2;
uint ty = (gid.y / 2) * 2;
uint tc = (gid.x % 2) * 2 + gid.y % 2;
uint tx = (x >> 1) << 1;
uint ty = (y >> 1) << 1;
uint tc = ((x % 2) << 1) + y % 2;
constexpr sampler sample(coord::pixel, filter::nearest, address::clamp_to_zero);
half4 inputs[16];
inputs[0] = inTexture.sample(sample, float2(tx - 1, ty - 1), tc);
inputs[1] = inTexture.sample(sample, float2(tx, ty - 1), tc);
inputs[2] = inTexture.sample(sample, float2(tx + 1, ty - 1), tc);
inputs[3] = inTexture.sample(sample, float2(tx + 2, ty - 1), tc);
inputs[4] = inTexture.sample(sample, float2(tx - 1, ty), tc);
inputs[5] = inTexture.sample(sample, float2(tx, ty), tc);
inputs[6] = inTexture.sample(sample, float2(tx + 1, ty), tc);
inputs[7] = inTexture.sample(sample, float2(tx + 2, ty), tc);
inputs[8] = inTexture.sample(sample, float2(tx - 1, ty + 1), tc);
inputs[9] = inTexture.sample(sample, float2(tx, ty + 1), tc);
inputs[10] = inTexture.sample(sample, float2(tx + 1, ty + 1), tc);
inputs[11] = inTexture.sample(sample, float2(tx + 2, ty + 1), tc);
inputs[12] = inTexture.sample(sample, float2(tx - 1, ty + 2), tc);
inputs[13] = inTexture.sample(sample, float2(tx, ty + 2), tc);
inputs[14] = inTexture.sample(sample, float2(tx + 1, ty + 2), tc);
inputs[15] = inTexture.sample(sample, float2(tx + 2, ty + 2), tc);
half4 base = biase[tc];
half4 res[4] = {base, base, base, base};
half f[3][3];
const uint kernelHXW = 9;
uint weightTo = tc * kernelHXW * 4;
uint weightTo = 4 * tc;
half4 res[4];
for (int c = 0; c < 4; ++c) {
for (int i = 0; i < 3; ++i) {
for (int j = 0; j < 3; ++j) {
f[i][j] = weights[weightTo++];
}
}
half I[16];
for (int i = 0; i < 16; ++i) {
I[i] = inputs[i][c];
}
half4x4 f = weights[weightTo + c];
half B[16];
half tmp1 = I[2] - I[10];
half tmp2 = I[1] - I[9];
half tmp2 = I[9] - I[1];
B[0] = I[0] - I[8] - tmp1;
B[1] = tmp1 + tmp2;
B[2] = tmp1 - tmp2;
B[3] = I[3] - I[11] - tmp2;
B[1] = tmp1 - tmp2;
B[2] = tmp1 + tmp2;
B[3] = I[3] - I[11] + tmp2;
tmp1 = I[6] + I[10];
tmp2 = I[5] + I[9];
B[4] = I[4] + I[8] - tmp1;
B[5] = tmp1 + tmp2;
B[6] = tmp1 - tmp2;
B[7] = I[7] + I[11] - tmp2;
tmp1 = I[6] - I[10];
tmp1 = I[10] - I[6];
tmp2 = I[5] - I[9];
B[8] = -I[4] + I[8] + tmp1;
B[9] = -tmp1 - tmp2;
B[10] = tmp2 - tmp1;
B[8] = I[8] - I[4] - tmp1;
B[9] = tmp1 - tmp2;
B[10] = tmp1 + tmp2;
B[11] = tmp2 - I[7] + I[11];
tmp1 = I[6] - I[14];
tmp1 = I[14] - I[6];
tmp2 = I[5] - I[13];
B[12] = -I[4] + I[12] + tmp1;
B[13] = -tmp1 - tmp2;
B[14] = tmp2 - tmp1;
B[12] = I[12] - I[4] - tmp1;
B[13] = tmp1 - tmp2;
B[14] = tmp1 + tmp2;
B[15] = tmp2 - I[7] + I[15];
half G[16];
G[0] = f[0][0];
G[1] = 0.5 * f[0][0] + 0.5 * f[0][1] + 0.5 * f[0][2];
G[2] = 0.5 * f[0][0] - 0.5 * f[0][1] + 0.5 * f[0][2];
G[3] = f[0][2];
G[4] = 0.5 * f[0][0] + 0.5 * f[1][0] + 0.5 * f[2][0];
G[5] = 0.25 * f[0][0] + 0.25 * f[0][1] + 0.25 * f[0][2] + 0.25 * f[1][0] + 0.25 * f[1][1] + 0.25 * f[1][2] + 0.25 * f[2][0] + 0.25 * f[2][1] + 0.25 * f[2][2];
G[6] = 0.25 * f[0][0] - 0.25 * f[0][1] + 0.25 * f[0][2] + 0.25 * f[1][0] - 0.25 * f[1][1] + 0.25 * f[1][2] + 0.25 * f[2][0] - 0.25 * f[2][1] + 0.25 * f[2][2];
G[7] = 0.5 * f[0][2] + 0.5 * f[1][2] + 0.5 * f[2][2];
G[8] = 0.5 * f[0][0] - 0.5 * f[1][0] + 0.5 * f[2][0];
G[9] = 0.25 * f[0][0] + 0.25 * f[0][1] + 0.25 * f[0][2] - 0.25 * f[1][0] - 0.25 * f[1][1] - 0.25 * f[1][2] + 0.25 * f[2][0] + 0.25 * f[2][1] + 0.25 * f[2][2];
G[10] = 0.25 * f[0][0] - 0.25 * f[0][1] + 0.25 * f[0][2] - 0.25 * f[1][0] + 0.25 * f[1][1] - 0.25 * f[1][2] + 0.25 * f[2][0] - 0.25 * f[2][1] + 0.25 * f[2][2];
G[11] = 0.5 * f[0][2] - 0.5 * f[1][2] + 0.5 * f[2][2];
G[12] = f[2][0];
G[13] = 0.5 * f[2][0] + 0.5 * f[2][1] + 0.5 * f[2][2];
G[14] = 0.5 * f[2][0] - 0.5 * f[2][1] + 0.5 * f[2][2];
G[15] = f[2][2];
half T[16];
for (int ii = 0; ii < 16; ++ii) {
T[ii] = B[ii] * G[ii];
}
T[0] = B[0] * f[0][0];
T[1] = B[1] * f[0][1];
T[2] = B[2] * f[0][2];
T[3] = B[3] * f[0][3];
T[4] = B[4] * f[1][0];
T[5] = B[5] * f[1][1];
T[6] = B[6] * f[1][2];
T[7] = B[7] * f[1][3];
T[8] = B[8] * f[2][0];
T[9] = B[9] * f[2][1];
T[10] = B[10] * f[2][2];
T[11] = B[11] * f[2][3];
T[12] = B[12] * f[3][0];
T[13] = B[13] * f[3][1];
T[14] = B[14] * f[3][2];
T[15] = B[15] * f[3][3];
tmp1 = T[1] + T[5] + T[9];
tmp2 = T[2] + T[6] + T[10];
res[0][c] += T[0] + T[4] + T[8] + tmp1 + tmp2;
res[1][c] += T[3] + T[7] + T[11] + tmp1 - tmp2;
res[0][c] = T[0] + T[4] + T[8] + tmp1 + tmp2;
res[1][c] = T[3] + T[7] + T[11] + tmp1 - tmp2;
tmp1 = T[5] - T[9] + T[13];
tmp2 = T[6] - T[10] + T[14];
res[2][c] += T[4] - T[8] + T[12] + tmp1 + tmp2;
res[3][c] += T[7] - T[11] + T[15] + tmp1 - tmp2;
res[2][c] = T[4] - T[8] + T[12] + tmp1 + tmp2;
res[3][c] = T[7] - T[11] + T[15] + tmp1 - tmp2;
}
outTexture.write(fmax(res[0], 0.0), uint2(tx, ty), tc);
outTexture.write(fmax(res[1], 0.0), uint2(tx + 1, ty), tc);
outTexture.write(fmax(res[2], 0.0), uint2(tx, ty + 1), tc);
outTexture.write(fmax(res[3], 0.0), uint2(tx + 1, ty + 1), tc);
}
//kernel void depthwise_conv_add_relu_3x3_half_winograd_naive(texture2d_array<half, access::sample> inTexture [[texture(0)]],
// texture2d_array<half, access::write> outTexture [[texture(1)]],
// constant MetalConvParam &param [[buffer(0)]],
// const device half *weights [[buffer(1)]],
// const device half4 *biase [[buffer(2)]],
// uint3 gid [[thread_position_in_grid]]) {
// uint ow = outTexture.get_width();
// uint oh = outTexture.get_height();
// if (gid.x >= ow || gid.y >= oh) {
// return;
// }
//
// uint tx = (gid.x / 2) * 2;
// uint ty = (gid.y / 2) * 2;
// uint tc = (gid.x % 2) * 2 + gid.y % 2;
//
// constexpr sampler sample(coord::pixel, filter::nearest, address::clamp_to_zero);
// half4 inputs[4][4];
// inputs[0][0] = inTexture.sample(sample, float2(tx - 1, ty - 1), tc);
// inputs[0][1] = inTexture.sample(sample, float2(tx, ty - 1), tc);
// inputs[0][2] = inTexture.sample(sample, float2(tx + 1, ty - 1), tc);
// inputs[0][3] = inTexture.sample(sample, float2(tx + 2, ty - 1), tc);
//
// inputs[1][0] = inTexture.sample(sample, float2(tx - 1, ty), tc);
// inputs[1][1] = inTexture.sample(sample, float2(tx, ty), tc);
// inputs[1][2] = inTexture.sample(sample, float2(tx + 1, ty), tc);
// inputs[1][3] = inTexture.sample(sample, float2(tx + 2, ty), tc);
//
// inputs[2][0] = inTexture.sample(sample, float2(tx - 1, ty + 1), tc);
// inputs[2][1] = inTexture.sample(sample, float2(tx, ty + 1), tc);
// inputs[2][2] = inTexture.sample(sample, float2(tx + 1, ty + 1), tc);
// inputs[2][3] = inTexture.sample(sample, float2(tx + 2, ty + 1), tc);
//
// inputs[3][0] = inTexture.sample(sample, float2(tx - 1, ty + 2), tc);
// inputs[3][1] = inTexture.sample(sample, float2(tx, ty + 2), tc);
// inputs[3][2] = inTexture.sample(sample, float2(tx + 1, ty + 2), tc);
// inputs[3][3] = inTexture.sample(sample, float2(tx + 2, ty + 2), tc);
//
// const uint kernelHXW = 9;
// uint weightTo = tc * kernelHXW * 4;
//
// half f[3][3];
//
// half4 base = biase[tc];
// half4 res[2][2];
// res[0][0] = base;
// res[0][1] = base;
// res[1][0] = base;
// res[1][1] = base;
//
// for (int c = 0; c < 4; ++c) {
// for (int i = 0; i < 3; ++i) {
// for (int j = 0; j < 3; ++j) {
// f[i][j] = weights[weightTo++];
// }
// }
// half I[4][4];
// for (int ii = 0; ii < 4; ++ii) {
// for (int jj = 0; jj < 4; ++jj) {
// I[ii][jj] = inputs[ii][jj][c];
// }
// }
// half B[4][4];
// B[0][0] = I[0][0] - I[0][2] - I[2][0] + I[2][2];
// B[0][1] = I[0][1] + I[0][2] - I[2][1] - I[2][2];
// B[0][2] = -I[0][1] + I[0][2] + I[2][1] - I[2][2];
// B[0][3] = -I[0][1] + I[0][3] + I[2][1] - I[2][3];
// B[1][0] = I[1][0] - I[1][2] + I[2][0] - I[2][2];
// B[1][1] = I[1][1] + I[1][2] + I[2][1] + I[2][2];
// B[1][2] = -I[1][1] + I[1][2] - I[2][1] + I[2][2];
// B[1][3] = -I[1][1] + I[1][3] - I[2][1] + I[2][3];
// B[2][0] = -I[1][0] + I[1][2] + I[2][0] - I[2][2];
// B[2][1] = -I[1][1] - I[1][2] + I[2][1] + I[2][2];
// B[2][2] = I[1][1] - I[1][2] - I[2][1] + I[2][2];
// B[2][3] = I[1][1] - I[1][3] - I[2][1] + I[2][3];
// B[3][0] = -I[1][0] + I[1][2] + I[3][0] - I[3][2];
// B[3][1] = -I[1][1] - I[1][2] + I[3][1] + I[3][2];
// B[3][2] = I[1][1] - I[1][2] - I[3][1] + I[3][2];
// B[3][3] = I[1][1] - I[1][3] - I[3][1] + I[3][3];
// half G[4][4];
// G[0][0] = f[0][0];
// G[0][1] = 0.5 * f[0][0] + 0.5 * f[0][1] + 0.5 * f[0][2];
// G[0][2] = 0.5 * f[0][0] - 0.5 * f[0][1] + 0.5 * f[0][2];
// G[0][3] = f[0][2];
// G[1][0] = 0.5 * f[0][0] + 0.5 * f[1][0] + 0.5 * f[2][0];
// G[1][1] = 0.25 * f[0][0] + 0.25 * f[0][1] + 0.25 * f[0][2] + 0.25 * f[1][0] + 0.25 * f[1][1] + 0.25 * f[1][2] + 0.25 * f[2][0] + 0.25 * f[2][1] + 0.25 * f[2][2];
// G[1][2] = 0.25 * f[0][0] - 0.25 * f[0][1] + 0.25 * f[0][2] + 0.25 * f[1][0] - 0.25 * f[1][1] + 0.25 * f[1][2] + 0.25 * f[2][0] - 0.25 * f[2][1] + 0.25 * f[2][2];
// G[1][3] = 0.5 * f[0][2] + 0.5 * f[1][2] + 0.5 * f[2][2];
// G[2][0] = 0.5 * f[0][0] - 0.5 * f[1][0] + 0.5 * f[2][0];
// G[2][1] = 0.25 * f[0][0] + 0.25 * f[0][1] + 0.25 * f[0][2] - 0.25 * f[1][0] - 0.25 * f[1][1] - 0.25 * f[1][2] + 0.25 * f[2][0] + 0.25 * f[2][1] + 0.25 * f[2][2];
// G[2][2] = 0.25 * f[0][0] - 0.25 * f[0][1] + 0.25 * f[0][2] - 0.25 * f[1][0] + 0.25 * f[1][1] - 0.25 * f[1][2] + 0.25 * f[2][0] - 0.25 * f[2][1] + 0.25 * f[2][2];
// G[2][3] = 0.5 * f[0][2] - 0.5 * f[1][2] + 0.5 * f[2][2];
// G[3][0] = f[2][0];
// G[3][1] = 0.5 * f[2][0] + 0.5 * f[2][1] + 0.5 * f[2][2];
// G[3][2] = 0.5 * f[2][0] - 0.5 * f[2][1] + 0.5 * f[2][2];
// G[3][3] = f[2][2];
// half T[4][4];
// for (int ii = 0; ii < 4; ++ii) {
// for (int jj = 0; jj < 4; ++jj) {
// T[ii][jj] = B[ii][jj] * G[ii][jj];
// }
// }
// half A[2][2];
// A[0][0] = T[0][0] + T[0][1] + T[0][2] + T[1][0] + T[1][1] + T[1][2] + T[2][0] + T[2][1] + T[2][2];
// A[0][1] = T[0][1] - T[0][2] + T[0][3] + T[1][1] - T[1][2] + T[1][3] + T[2][1] - T[2][2] + T[2][3];
// A[1][0] = T[1][0] + T[1][1] + T[1][2] - T[2][0] - T[2][1] - T[2][2] + T[3][0] + T[3][1] + T[3][2];
// A[1][1] = T[1][1] - T[1][2] + T[1][3] - T[2][1] + T[2][2] - T[2][3] + T[3][1] - T[3][2] + T[3][3];
// for (int i = 0; i < 2; ++i) {
// for (int j = 0; j < 2; ++j) {
// res[i][j][c] += A[i][j];
// }
// }
// }
//
// for (int i = 0; i < 2; ++i) {
// for (int j = 0; j < 2; ++j) {
// half4 output = fmax(res[i][j], 0.0);
// outTexture.write(output, uint2(tx + j, ty + i), tc);
// }
// }
//}
half4 base = biase[tc];
outTexture.write(fmax(res[0] + base, 0.0), uint2(tx, ty), tc);
outTexture.write(fmax(res[1] + base, 0.0), uint2(tx + 1, ty), tc);
outTexture.write(fmax(res[2] + base, 0.0), uint2(tx, ty + 1), tc);
outTexture.write(fmax(res[3] + base, 0.0), uint2(tx + 1, ty + 1), tc);
}
kernel void conv_add_relu_5x1_half(texture2d_array<half, access::sample> inTexture [[texture(0)]],
texture2d_array<half, access::write> outTexture [[texture(1)]],
......
metal/paddle-mobile/paddle-mobile/Src/Framework/Tensor.swift
浏览文件 @ 64aa8f05
......@@ -36,6 +36,10 @@ class DataConverter<P: PrecisionProtocol> {
func getToDim(fromDim: Dim, layout: DataLayout) -> (dim: Dim, layout: DataLayout) {
fatalError(" need imp")
}
func capacity(fromDim: Dim) -> Int? {
return nil
}
}
/// [ outputChannels ][ inputChannels ][ kernelHeight ][ kernelWidth ] ->
......@@ -81,6 +85,119 @@ class MPSPointerConverter<P: PrecisionProtocol>: DataConverter<P>{
}
}
class WinogradPointerConverter<P: PrecisionProtocol>: DataConverter<P>{
override func convert(from: UnsafeMutablePointer<P>, to: UnsafeMutablePointer<P>, fromDim: Dim) {
let N = fromDim[0]
let C = fromDim[1]
let H = fromDim[2]
let W = fromDim[3]
if H != 3 || W != 3 {
fatalError("not support")
}
for n in 0..<N {
for c in 0..<C {
let fromOffset = n * C * H * W + c * H * W
let toOffset = n * C * (H + 1) * (W + 1) + c * (H + 1) * (W + 1)
func f(_ h: Int, _ w: Int) -> P {
return from[fromOffset + h * W + w]
}
let c05 = P(Float(0.5))
let c025 = P(Float(0.25))
to[toOffset] = f(0, 0);
to[toOffset + 1] = c05 * f(0, 0)
to[toOffset + 1] = to[toOffset + 1] + c05 * f(0, 1)
to[toOffset + 1] = to[toOffset + 1] + c05 * f(0, 2)
to[toOffset + 2] = c05 * f(0, 0)
to[toOffset + 2] = to[toOffset + 2] - c05 * f(0, 1)
to[toOffset + 2] = to[toOffset + 2] + c05 * f(0, 2)
to[toOffset + 3] = f(0, 2)
to[toOffset + 4] = c05 * f(0, 0)
to[toOffset + 4] = to[toOffset + 4] + c05 * f(1, 0)
to[toOffset + 4] = to[toOffset + 4] + c05 * f(2, 0)
to[toOffset + 5] = c025 * f(0, 0)
to[toOffset + 5] = to[toOffset + 5] + c025 * f(0, 1)
to[toOffset + 5] = to[toOffset + 5] + c025 * f(0, 2)
to[toOffset + 5] = to[toOffset + 5] + c025 * f(1, 0)
to[toOffset + 5] = to[toOffset + 5] + c025 * f(1, 1)
to[toOffset + 5] = to[toOffset + 5] + c025 * f(1, 2)
to[toOffset + 5] = to[toOffset + 5] + c025 * f(2, 0)
to[toOffset + 5] = to[toOffset + 5] + c025 * f(2, 1)
to[toOffset + 5] = to[toOffset + 5] + c025 * f(2, 2)
to[toOffset + 6] = c025 * f(0, 0)
to[toOffset + 6] = to[toOffset + 6] - c025 * f(0, 1)
to[toOffset + 6] = to[toOffset + 6] + c025 * f(0, 2)
to[toOffset + 6] = to[toOffset + 6] + c025 * f(1, 0)
to[toOffset + 6] = to[toOffset + 6] - c025 * f(1, 1)
to[toOffset + 6] = to[toOffset + 6] + c025 * f(1, 2)
to[toOffset + 6] = to[toOffset + 6] + c025 * f(2, 0)
to[toOffset + 6] = to[toOffset + 6] - c025 * f(2, 1)
to[toOffset + 6] = to[toOffset + 6] + c025 * f(2, 2)
to[toOffset + 7] = c05 * f(0, 2)
to[toOffset + 7] = to[toOffset + 7] + c05 * f(1, 2)
to[toOffset + 7] = to[toOffset + 7] + c05 * f(2, 2)
to[toOffset + 8] = c05 * f(0, 0)
to[toOffset + 8] = to[toOffset + 8] - c05 * f(1, 0)
to[toOffset + 8] = to[toOffset + 8] + c05 * f(2, 0)
to[toOffset + 9] = c025 * f(0, 0)
to[toOffset + 9] = to[toOffset + 9] + c025 * f(0, 1)
to[toOffset + 9] = to[toOffset + 9] + c025 * f(0, 2)
to[toOffset + 9] = to[toOffset + 9] - c025 * f(1, 0)
to[toOffset + 9] = to[toOffset + 9] - c025 * f(1, 1)
to[toOffset + 9] = to[toOffset + 9] - c025 * f(1, 2)
to[toOffset + 9] = to[toOffset + 9] + c025 * f(2, 0)
to[toOffset + 9] = to[toOffset + 9] + c025 * f(2, 1)
to[toOffset + 9] = to[toOffset + 9] + c025 * f(2, 2)
to[toOffset + 10] = c025 * f(0, 0)
to[toOffset + 10] = to[toOffset + 10] - c025 * f(0, 1)
to[toOffset + 10] = to[toOffset + 10] + c025 * f(0, 2)
to[toOffset + 10] = to[toOffset + 10] - c025 * f(1, 0)
to[toOffset + 10] = to[toOffset + 10] + c025 * f(1, 1)
to[toOffset + 10] = to[toOffset + 10] - c025 * f(1, 2)
to[toOffset + 10] = to[toOffset + 10] + c025 * f(2, 0)
to[toOffset + 10] = to[toOffset + 10] - c025 * f(2, 1)
to[toOffset + 10] = to[toOffset + 10] + c025 * f(2, 2)
to[toOffset + 11] = c05 * f(0, 2)
to[toOffset + 11] = to[toOffset + 11] - c05 * f(1, 2)
to[toOffset + 11] = to[toOffset + 11] + c05 * f(2, 2)
to[toOffset + 12] = f(2, 0)
to[toOffset + 13] = c05 * f(2, 0)
to[toOffset + 13] = to[toOffset + 13] + c05 * f(2, 1)
to[toOffset + 13] = to[toOffset + 13] + c05 * f(2, 2)
to[toOffset + 14] = c05 * f(2, 0)
to[toOffset + 14] = to[toOffset + 14] - c05 * f(2, 1)
to[toOffset + 14] = to[toOffset + 14] + c05 * f(2, 2)
to[toOffset + 15] = f(2, 2)
}
}
}
override func getToDim(fromDim: Dim, layout: DataLayout) -> (dim: Dim, layout: DataLayout) {
if layout != DataLayout.NCHW() {
fatalError("not support")
}
let N = fromDim[0]
let C = fromDim[1]
let H = fromDim[2]
let W = fromDim[3]
if H != 3 || W != 3 {
fatalError("not support")
}
let toDim = Dim.init(inDim: [N, C, H + 1, W + 1])
return (dim: toDim, layout: DataLayout.NCHW())
}
override func capacity(fromDim: Dim) -> Int? {
let N = fromDim[0]
let C = fromDim[1]
let H = fromDim[2]
let W = fromDim[3]
if H != 3 || W != 3 {
fatalError("not support")
}
return N * C * (H + 1) * (W + 1)
}
}
class Tensor<P: PrecisionProtocol>: Tensorial {
var data: Data
......@@ -135,9 +252,10 @@ class Tensor<P: PrecisionProtocol>: Tensorial {
}
func convert(converter: DataConverter<P>) -> UnsafeMutablePointer<P> {
let to = UnsafeMutablePointer<P>.allocate(capacity: numel())
let toCapacity = converter.capacity(fromDim: dim) ?? numel()
let to = UnsafeMutablePointer<P>.allocate(capacity: toCapacity)
converter.convert(from: data.pointer, to: to, fromDim: dim)
data = Data.init(inCount: numel(), inPointer: to)
data = Data.init(inCount: toCapacity, inPointer: to)
let dimAndLayout = converter.getToDim(fromDim: dim, layout: layout)
dim = dimAndLayout.dim
layout = dimAndLayout.layout
......
metal/paddle-mobile/paddle-mobile/Src/Operators/Kernels/ConvAddKernel.swift
浏览文件 @ 64aa8f05
......@@ -194,6 +194,9 @@ class ConvAddKernel<P: PrecisionProtocol>: Kernel, Computable {
let inMetalParam = MetalConvParam.init(offsetX: Int16(offsetX), offsetY: Int16(offsetY), offsetZ: Int16(offsetZ), strideX: UInt16(param.stride[0]), strideY: UInt16(param.stride[1]), dilationX: UInt16(param.dilations[0]), dilationY: UInt16(param.dilations[1]))
metalParam = inMetalParam
if type(of: self).isWinoGrad(functionName: functionName) {
let _ = param.filter.convert(converter: WinogradPointerConverter<P>.init())
}
let padWhenOneC = !(param.filter.channel == 1 && param.filter.n == param.input.tensorDim[1])
param.filter.initBuffer(device: device, precision: GlobalConfig.shared.computePrecision, padWhenOneC: padWhenOneC)
......
metal/paddle-mobile/paddle-mobile/Src/Operators/Kernels/ConvAddReluKernel.swift
浏览文件 @ 64aa8f05
......@@ -15,7 +15,7 @@ class ConvAddReluKernel<P: PrecisionProtocol>: ConvAddKernel<P> {
if param.filter.width == 1 && param.filter.height == 1 {
return "conv_add_relu_1x1_half"
} else if param.filter.channel == 1 && param.filter.n == param.input.tensorDim[1] {
if param.filter.n == 16 && param.stride[0] == 1 && param.stride[1] == 1 && param.input.tensorDim[2] % 2 == 0 && param.input.tensorDim[3] % 2 == 0 && false {
if param.filter.width == 3 && param.filter.height == 3 && param.stride[0] == 1 && param.stride[1] == 1 && param.filter.n == 16 {
return "depthwise_conv_add_relu_3x3_half_winograd"
} else {
return "depthwise_conv_add_relu_3x3_half"
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或